English High Court Decision: AI Inventions are Patentable

The High Court of Justice in London recently issued a judgment on the question of whether or not an artificial intelligence invention falls within the ambit of “a program for a computer … as such” and therefore not patentable subject matter pursuant to section 1(2)(c) Patents Act 1977.

By the use of a trained artificial neural network (ANN) the system of the proposed patent learns someone’s tastes and is able to offer suggestions of similar files (for example, music or art) in terms of human perception and emotion irrespective of the genre of the music.

At the U.K. Intellectual Property Office (UKIPO) the patent application filed by Emotional Perception AI Ltd. was refused on the basis that it did. Emotional Perception appealed that decision to the High Court. This article considers the judgment of the court, the full version of which is available here. The UKIPO’s role in such proceedings is to justify its earlier decision and to represent the public interest by putting forward to the court arguments to challenge those made by the patent applicant.

As was established in Aerotel Ltd. v Telco Holdings Ltd. [2007] RPC 7, there are four steps to be followed when considering whether an invention falls foul of the exclusion:

1. Properly construe the claim.
2. Identify the actual contribution (although at the application stage this might have to be the alleged contribution).
3. Ask whether it falls solely within the excluded matter.
4. If the third step has not covered it, check whether the actual or alleged contribution is actually technical.

With regards to the first and second Aerotel questions, the UKIPO considered, and the applicant did not disagree, that the contribution in this case was:

“[A] method in which an ANN is trained on pairs of modally identical files (for example pairs of songs) in order to map the distance between their property vectors in property space towards the distance between their semantic vectors in semantic space. When an input (for example representative of a user selected song) is presented to the ANN, it generates a file vector in property space for said input. This file vector is then compared to the property vectors of reference files (representative of a library of songs, say) to identify those files having similar vectors in property space to that of the input. Such files will be semantically similar to the input by virtue of the ANN having been trained to map/converge distances in property space towards distances in semantic space (i.e. so that it generates similar property vectors for semantically similar files). A file can then be sent to, and output by, a user device. In this way, it provides a tool for recommending semantically similar files."

It was held by the Hearing Officer at the UKIPO that the ANN could be implemented in hardware or software; a point which was not disputed on appeal. A hardware ANN consists of neurons arranged in layers. Each neuron processes signals and produces an output which is passed to neurons in other layers, save for the last layer which produces an output from the system. The processing is done according to internal instructions and further processes such as weights and biases applied by the neurons. The software or emulated ANN runs in software an emulation of the hardware ANN.

The judge addressed the third Aerotel question by posing to the parties these questions: “what is the computer?” and “where is the program?” He turned first to consider whether a hardware ANN was a computer. He started by considering the Oxford English Dictionary definition of a computer:

“An electronic device (or system of devices) which is used to store, manipulate, and communicate information, perform complex calculations, or control or regulate other devices or machines, and is capable of receiving information (data) and of processing it in accordance with variable procedural instructions (programs or software); esp. a small, self-contained one for individual use in the home or workplace, used esp. for handling text, images, music, and video, accessing and using the internet, communicating with other people (e.g. by means of email), and playing games.”

He then went on to say:

“The first part of that definition is capable of describing a hardware ANN. The ‘variable procedural instructions’ are, while it is learning, the elements by which it learns and back-propagates, and the frozen state contains biases, weighting and so on which it has learnt for itself and which one might call instructions. However, the key is that it is processing data. I consider that in everyday parlance it would be regarded as a computer, and ought to be treated as one within the exclusion.”

As far as a software emulation of an ANN is concerned the judge had no difficulty in concluding that a computer is involved.

The judge then went on to consider the question of the program. The UKIPO conceded that in the case of a hardware ANN, there is no program and therefore if the patent were limited to a hardware ANN the question of whether the exclusion is invoked does not arise. The judge seemingly was disappointed not to engage with this point; he writes:

“I am not myself minded to consider the correctness of the concession, though I think that the debate would have been interesting had the concession not been made.”

The UKIPO’s case was that in the emulated ANN there were two computer programs engaged: at the training stage when a program was necessary to provide the training, and secondly the program which runs the emulation. As to the latter the judge concluded:

“It seems to me that it is appropriate to look at the emulated ANN as, in substance, operating at a different level (albeit metaphorically) from the underlying software on the computer, and it is operating in the same way as the hardware ANN. If the latter is not operating a program then neither is the emulation.”

As to the former, the judge stated:

“I therefore turn to consider whether the presence of a computer program at the training stage means that the patent claims ‘a program for a computer … as such.’ The programming involves setting the training objectives in terms of the structure of the ANN (if in software) and the training objectives. It is not possible to define the programming any further than that. … So far, then, there is a computer program involved, at least at that level. However, it does not seem to me that the [patent] claim claims that program. What is said to be special is the idea of using pairs of files for training, and setting the training objective and parameters accordingly. If that is right, and I consider it is, then the actual program is a subsidiary part of the claim and is not what is claimed. The claims go beyond that. The idea of the parameters itself is not necessarily part of the program. On this footing as a matter of construction the claim is not to a computer program at all. The exclusion is not invoked.”

Finally, applying the belt-and-braces check that is the fourth Aerotel question (i.e., whether the invention is actually technical), the judge recorded:

“The correct view of what happened, for these purposes, is that a file has been identified, and then moved, because it fulfilled certain criteria. True it is that those criteria are not technical criteria in the sense that they can be described in purely technical terms, but they are criteria nonetheless, and the ANN has certainly gone about its analysis and selection in a technical way. It is not just any old file; it is a file identified as being semantically similar by the application of technical criteria which the system has worked out for itself. So the output is of a file that would not otherwise be selected. That seems to me to be a technical effect outside the computer for these purposes, and when coupled with the purpose and method of selection it fulfils the requirement of technical effect in order to escape the exclusion. I do not see why the possible subjective effect within a user's own non-artificial neural network should disqualify it for these purposes. … the invention is not just one depending on the effect of the computerised process on the user. There is more than that. There is a produced file with (it is said) certain attributes. The file produced then goes on to have an effect on the user (if the thing works at all) but one cannot ignore the fact that a technical thing is actually produced. It would not matter if the user never listened to the file. The file, with its similarity characteristics, is still produced via the system which has set up the identification system and then implemented it.” 

Thus, the court held that the exclusion was not invoked — such an AI system is not “a program for a computer … as such.” The UKIPO has subsequently issued guidance (available here) confirming that:

“Patent examiners should not object to inventions involving ANNs under the ‘program for a computer’ exclusion.”

A decade or more ago there was a general misconception that software implemented inventions could not be patented because they would fall within the computer program exemption. Fortunately that misunderstanding is now largely in the past and businesses are generally alert to the possibility of obtaining patent protection for software implemented inventions. 

This judgment should provide similar clarity and encouragement to those inventing in the AI space. That said, the judgment is in some respects disappointing. The UKIPO seemed to change its position at various stages and both parties made concessions which, reading the judgment, one gets the impression the judge was dubious about. It does, however, set out the law in the U.K. on this point, which will stand unless and until it is overturned, perhaps in a revocation action of a granted patent.